Modeling climate change impact on inflow and hydropower generation of Nangbeto dam in West Africa using multi-model CORDEX ensemble and ensemble machine learning

Climate change (CC) poses a threat to renewable hydropower, which continues to play a significant role in energy generation in West Africa (WA). Thus, the assessment of the impacts of climate change and climate variability on hydropower generation is critical for dam management. This study develops a framework based on ensemble climate models and ensemble machine learning methods to assess the projected impacts of CC on inflow to the reservoir and hydropower generation at the Nangbeto Hydropower plant in WA. Inflow to reservoir and energy generation for the future (2020–2099) is modeled using climate models output data from Coordinated Regional Downscaling Experiment to produce a publicly accessible hydropower dataset from 1980 to 2099. The bias-adjusted ensemble mean of eleven climate models for representative concentration pathways (RC4.5 and RCP8.5) are used. The added value of this approach is to use fewer input data (temperature and precipitation) while focusing on their lagged effect on inflow and energy. Generally, the model output strongly correlates with the observation (1986–2005) with a Pearson correlation of 0.86 for energy and 0.82 for inflow while the mean absolute error is 2.97% for energy and 9.73% for inflow. The results reveals that both inflow and energy simulated over the future periods (2020–2039, 2040–2059, 2060–2079, and 2080–2099) will decrease relative to the historical period (1986–2005) for both RCPs in the range of (2.5–20.5% and 1–8.5% for inflow and energy, respectively), at annual, monthly and seasonal time scales. Therefore, these results should be considered by decision-makers when assessing the best option for the energy mix development plan.